TLDR: Like an LLM, the input is a string, but the output is not a completion. The output is a ranking of elements from a certain enum (e.g, [Yes/No], [A/B/C/D]). They use a technique called Reinforcement Learning for Calibrated Decisions (RLCD) instead of RLHF. Also, inference is a lot faster.

https://docs.typesafe.ai/primitives has a much better explanation

Great! Can't wait until a few billion dollars are wasted on a new Meta product which becomes a nothingburger even before it starts

> I didn't realize that open math problems were a finite resource.

That is exactly what Tao is explaining in that tweet.

TLDR: Open Problems are infinite, but those which are at the boundary of easy and hard problems and are interesting are far more scarce

I read the first sentence and thought: what is the government of Albania doing in Australia?

OMG, I've got it now

"Albanese" is the second name of the Australian Prime Minister

Albanese is an Italian name, for the record

In Albania, Albania is called Shqipëria.

I was thinking albanese were the people of Albania (I know it's albanian, but got confuse for a moment)

Yep, I spent a hot minute puzzling over that sentence

Is it related to reticulum?

Reticulum can use it as one of several transport layers. LoRa is just layer 1 only

Are H1Bs still being processed? They are dual intent visas as far as I know

Yes, they are. Immigrant visa = green card. Dual intent just means you are allowed to apply for a green card.

So only PR applications are affected?

I recommend Russ Cox's articles on implementing a regex engine: https://swtch.com/~rsc/regexp/

It is very relevant

Related:

Don't Be a Meat Proxy: https://gruhn.me/blog/2026-08-03/

No Meat Proxy: https://nomeatproxy.com/

How to Passive-Aggressively Shame People Who Use LLMs Selfishly: https://joshmoody.org/blog/selfish-ai/

AI;DR: https://www.rickmanelius.com/p/aidr-ai-didnt-read

AI;DR https://news.ycombinator.com/item?id=49336573

"to power the next wave of GDP growth" is a wonderfully shallow but impressive clause.

I didn't know Linear has "AI features". Linear is boring, but that's actually fine by me.

I use LLMs to write my code, but this does not show up in this data.